Every article argues for something. Most of them also argue against something — and the thing they are fighting is rarely named directly. The Adversarial Distiller is a four-persona method for finding both: the incontrovertible truth that survives every attack, and the opposing claim the article is really trying to defeat.
The output is three things: a main truth, an anti-truth, and a ranked set of subsidiary truths. These are not summaries. A summary tells you what the article says. A distillation tells you what the article cannot escape saying, and what it would have to concede if it were wrong.
Workflow
The four personas run in sequence. Each one constrains the next. The Witness cannot invent — it can only observe what the prior two exposed. The Cartographer cannot elevate a claim the Witness did not confirm.
The Four Personas
The Advocate argues for the article as if their reputation depends on it. They identify the two or three strongest supporting moves, state what the article gets undeniably right, and conclude with the claim they cannot shake. The Advocate is not a cheerleader — they are the best possible version of the article's own argument.
The Challenger finds every weakness, false assumption, and gap. They identify the most vulnerable premise, surface the one thing asserted without sufficient evidence, and state the strongest counter-claim a rigorous opponent would make. The Challenger is not a contrarian — they are making the most honest case against the article that can be made.
The Witness does not advocate. They watched the debate and report what neither side could touch. The key question: what did the Challenger have to concede, even while attacking? What did the Advocate claim that the Challenger could not actually disprove? The overlap is the incontrovertible zone. The Witness applies a collapse test: which surviving claim, if removed, would make the article's core argument fall apart? That is the main truth.
The Cartographer maps what else is true. From the debris of the debate, they identify subsidiary claims that support the main truth without being the main truth itself — ranked by how much load-bearing weight they carry. They also name the anti-truth: the sharpest opposing claim that would be true if the main truth were false. If the anti-truth does not make the author slightly uncomfortable, it is not sharp enough.
Running the Method
The Distiller can be run in three ways.
With an AI assistant. Brief the model with the article text and ask it to run each persona in sequence, showing its reasoning before outputting the final three results. The value of showing reasoning is that it lets you audit the persona logic — if the Challenger missed an obvious weakness, you want to see why. Running all four personas in a single prompt tends to collapse them; running them as a sequence, with each output informing the next, produces sharper results.
In a small group. Assign one persona to each participant. The Advocate and Challenger argue their positions aloud; the Witness synthesises; the Cartographer maps. This works particularly well for high-stakes documents — strategy papers, policy proposals, investment theses — where the cost of a weak argument going public is significant.
Solo, in writing. Write each persona's analysis before moving to the next. Do not skip to the Witness before both the Advocate and Challenger are fully argued. The temptation is to jump straight to the conclusion. The Witness conclusion is only as good as the argument that preceded it.
The Collapse Test
The most important single step in the method. After the Witness has identified what survived, ask: which of these claims, if removed, would make the rest of the article's argument fall apart?
This is not asking what the article is about. It is asking what the article depends on. An article can be about many things but depend on one thing. That one thing is the main truth.
If you cannot identify a single claim that passes this test, the article does not yet have a spine. That is a finding worth having before publication.
The Anti-Truth Standard
The anti-truth must sting. It should be the claim that a thoughtful, well-intentioned person who disagreed with the article would make — not a strawman, not a dismissal, but the best opposing argument available.
A useful test: would the author of the article be willing to include the anti-truth as a named objection and address it directly? If yes, it is good enough. If the author can dismiss it in a sentence, it is not sharp enough.
The anti-truth is not a rebuttal of the article. It is an honest map of what the article has not yet answered.
Benefits
Risks
Mitigations
Handoff
The Adversarial Distiller produces three outputs: a one-sentence main truth, a one-sentence anti-truth, and a ranked list of subsidiary truths. These can be used directly as metadata for content management, as the spine of a response or follow-up piece, or as a briefing summary for anyone who needs to understand an argument quickly without reading the full article.
In a publishing workflow, the distillation outputs sit alongside the article itself and are available to any downstream process — distribution, response tracking, or further research — that needs a structured representation of what the article claims.